Swarm Convergence on a Maze

When the floor does the thinking.

No ant knows where the exit is or how to reach it. Each only senses the floor beside it and lays a trail scaled by its own success — yet the colony collectively finds the exit and then shortens the route, ratcheting toward the near-optimal path. Stigmergy doing graph search.

Best route–
Optimal–
First found–
Solves0
attractive trail "wandered, nothing here" proven ant still searching
Follow trails:
Why this, and not just foraging? A lone forager (Pheromone Foraging) finds food on raw local trails but can't tell a good route from a bad one. Add success-scaling and authority and the floor gains quality control: dozens of mediocre searchers collectively find a route and certify it, then ratchet it shorter — pooled, self-correcting search no single agent can do. That's the leap from emergence to emergent optimisation.

How it works — earned, local stigmergy. An ant that hasn't solved yet lays mild repulsion as it wanders ("nothing useful here"), so the unproven swarm disperses. The instant one reaches the exit it reinforces its whole route with attraction scaled by success — and success is 1/length, so a shorter solve lays a stronger trail. Evaporation erodes whatever isn't re-walked, so long detours fade and the colony slides onto shorter routes. Authority gating: an ant follows a trail only if it was laid by someone who out-scored it — you don't herd after a peer, only after a proven better. Flip it to "any trail" to watch the swarm chase itself onto worse paths.

Tie to Neuron Lab. This is the project's signed swarmCrumbs floor exactly: deposits scaled by the depositor's score, ant-style evaporation, and a per-cell swarmCrumbScore authority a reader must beat to follow — "the swarm disperses while searching, then converges onto the first trail that actually scores," and ratchets shorter from there. All local and earned: no beacon, no broadcast of where the exit is.